Open access, freely available online PLoS MEDICINE Rethinking the ‘‘ of Affluence’’ Paradigm: Global Patterns of Nutritional Risks in Relation to Economic Development

Majid Ezzati1*, Stephen Vander Hoorn2, Carlene M. M. Lawes2, Rachel Leach3, W. Philip T. James3, Alan D. Lopez4, Anthony Rodgers2, Christopher J. L. Murray1 1 Harvard School of Public , Boston, Massachusetts, United States of America, 2 Clinical Trials Research Unit, University of Auckland, New Zealand, 3 International Task Force, London, United Kingdom, 4 School of Population Health, University of Queensland, Brisbane, Australia

Competing Interests: The authors have declared that no competing ABSTRACT interests exist. Background Author Contributions: ME, SV, and CJLM designed the study. SV, Cardiovascular diseases and their nutritional risk factors—including overweight and obesity, CMML, RL, and WPTJ collected data. ME, SV, RL, and WPTJ elevated blood pressure, and cholesterol—are among the leading causes of global mortality analyzed the data. ME, SV, CMML, and morbidity, and have been predicted to rise with economic development. RL, WPTJ, ADL, AR, and CJLM contributed to writing the paper. Methods and Findings Academic Editor: Thomas No- votny, University of California at We examined age-standardized mean population levels of body mass index (BMI), systolic San Francisco, United States of America blood pressure, and total cholesterol in relation to national income, food share of household expenditure, and in a cross-country analysis. Data were from a total of over 100 Citation: Ezzati M, Vander Hoorn countries and were obtained from systematic reviews of published literature, and from national S, Lawes CMM, Leach R, James WPT, et al. (2005) Rethinking the and international health agencies. ‘‘diseases of affluence’’ paradigm: BMI and cholesterol increased rapidly in relation to national income, then flattened, and Global patterns of nutritional risks in relation to economic develop- eventually declined. BMI increased most rapidly until an income of about I$5,000 (international ment. PLoS Med 2(5): e133. dollars) and peaked at about I$12,500 for females and I$17,000 for males. Cholesterol’s point of inflection and peak were at higher income levels than those of BMI (about I$8,000 and I$18,000, Received: December 3, 2004 respectively). There was an inverse relationship between BMI/cholesterol and the food share of Accepted: March 7, 2005 Published: May 3, 2005 household expenditure, and a positive relationship with proportion of population in urban areas. Mean population blood pressure was not correlated or only weakly correlated with the DOI: economic factors considered, or with cholesterol and BMI. 10.1371/journal.pmed.0020133

Copyright: Ó 2005 Ezzati et al. This Conclusions is an open-access article distrib- uted under the terms of the Crea- tive Commons Attribution License, When considered together with evidence on shifts in income–risk relationships within which permits unrestricted use, developed countries, the results indicate that cardiovascular risks are expected to distribution, and reproduction in systematically shift to low-income and middle-income countries and, together with the any medium, provided the original work is properly cited. persistent burden of infectious diseases, further increase inequalities. Preventing obesity should be a priority from early stages of economic development, accompanied by Abbreviations: BMI, body mass population-level and personal interventions for blood pressure and cholesterol. index; SBP, systolic blood pressure; WHO, World Health Organization

*To whom correspondence should be addressed. E-mail: mezzati@ hsph.harvard.edu

PLoS Medicine | www.plosmedicine.org0404 May 2005 | Volume 2 | Issue 5 | e133 Global Patterns of Nutritional Risks

Figure 1. Global Mortality and Burden of Disease Attributable to Cardiovascular Diseases and Their Major Risk Factors for People 30 y of Age and Older The size of each circle is proportional to the number of deaths (left) or burden of disease (right; measured in disability-adjusted life years) (in millions). Overweight and obesity affect non-cardiovascular diseases, including , endometrial and colon , post-menopausal breast , and osteoarthritis, shown as the portions of yellow circles that fall outside the circle [57]. The mortality estimates exclude osteoarthritis, which results in morbidity but not direct deaths. Disease burden does include nonfatal health outcomes associated with diabetes and osteoarthritis (hence the larger size of the circle for overweight and obesity relative to those for blood pressure and cholesterol). Source: re-analysis of data from Ezzati et al. [57,58]. DOI: 10.1371/journal.pmed.0020133.g001

Introduction systematically examined the population-level relationships between three leading nutritional cardiovascular risk fac- Cardiovascular diseases and their nutritional risk factors tors—overweight and obesity, elevated blood pressure, and are among the leading causes of mortality and morbidity cholesterol—and three economic variables using data for globally (Figure 1), and have been predicted to rise over the over 100 countries. Analysis of multiple nutritional risks next few decades [1–3]. Aging of the world’s population is a shows more complex economic–epidemiological patterns key driver of the expected increase, because cardiovascular than those predicted by simple descriptions such as the disease rates tend to increase with age. In addition to this ‘‘diseases of affluence’’ or ‘‘Western disease’’ paradigms. More demographic change, an epidemiological change that in- importantly, focusing on multiple risk factors helps identify volves increases in age-specific rates of cardiovascular specific intervention and policy options and priorities, with diseases in developing countries has also been predicted in implications for societies at various levels of economic some analyses [4]. This epidemiological change is a corollary development. to a predicted population-wide rise in cardiovascular disease risk factors including obesity, blood pressure, cholesterol, and Methods use with increasing income, originally referred to as the ‘‘diseases of affluence’’ or ‘‘Western disease’’ paradigm We examined the cross-sectional relationship between [5,6]. A number of challenges have been made to the diseases mean population blood pressure, cholesterol, and body mass of affluence paradigm. For example, it has been observed that index (BMI) and three socioeconomic variables: national cardiovascular diseases and some of their risk factors (e.g., income, average share of household expenditure spent on smoking) may decline once they have peaked [7]. It has also food, and proportion of population in urban areas. Blood been documented that within upper-middle-income and pressure, cholesterol, and BMI are well-established cardio- high-income countries, cardiovascular diseases and risk vascular risk factors and provide aggregate indicators of more factors are increasingly concentrated among the lowest complex dietary patterns (e.g., caloric intake, and consump- socioeconomic groups [8–11]. tion of salt, fats of different composition, and fruits and Despite these challenges in specific populations, at the vegetables) and physical activity. Further, there are more global level, predictions about rising levels of cardiovascular comparable data from population-based health and risk factors with economic development continue to be made surveys on these physiological indicators than on dietary [2,12,13]. The global health aspect of the diseases of affluence patterns and physical activity, because these indicators can be paradigm is particularly important because it implies that a more easily defined in a consistent manner and measured large proportion of the world’s population, who live in using standard techniques. National income is the commonly middle-income countries, will soon face both aging popula- used indicator for a society’s material well-being. The share of tions and rising age-specific cardiovascular disease rates, and household expenditure spent on food is a measure of how will require increasing focus on policies and interventions to household economic resources may constrain food purchase. reduce the resulting disease burden [13]. Yet the timing of If food forms a large proportion of total household initiating interventions during a society’s economic develop- expenditure, households may limit the total amount of food ment and the specific form of the required interventions have consumed, which should result in lower obesity and possibly not been addressed based on systematic analyses of lower levels of other risk factors if they are affected by and disease profiles. The diseases of affluence paradigm also overconsumption, or households may switch to less expensive implies that cardiovascular disease risk factors are not urgent but lower quality foods that increase various nutritional risk public health concerns for low-income populations. We factors (e.g., higher salt, leading to higher blood pressure, or

PLoS Medicine | www.plosmedicine.org0405 May 2005 | Volume 2 | Issue 5 | e133 Global Patterns of Nutritional Risks

Table 1. Risk and Socioeconomic Variables Used in the Analysis

Risk and Economic Indicator Variable Number of Data Sources Factors Countries with Data

Blood pressure (age  30) Level of SBP 85 Systematic review of more than 200 studies with over 700,000 participants [17] updated with WHO non-communicable risk factor surveillance report [59] Cholesterol (age  30) Level of total 64 Systematic review of 150 surveys with over blood cholesterol 550,000 participants [18] updated with WHO report on non-communicable risk factors [59] BMI (age  30) BMI (height over 69 Systematic review of published literature, weight squared) WHO, and other reports, and requests to ministries of health and researchers [57,19] National income, adjusted Per-capita gross All Penn World Table [60] and WHO for purchasing power domestic product in international dollarsa Average share of household expenditure spent Percent household 71 National household surveys [61] on food expenditure Proportion of population Percent All United Nations Population Division [62] in urban areas population

a International dollar (Int$) is adjusted for purchasing power, and for inflation because the years of data for BMI, SBP, and cholesterol varied across countries. DOI: 10.1371/journal.pmed.0020133.t001

higher and fat, leading to higher obesity) [14]. The ness [17–19]. There were, nonetheless, differences among proportion of the population living in urban areas is a proxy studies in a number of dimensions: sample size, national indicator of a number of environmental and lifestyle (versus sub-national) representativeness, year of the study, variables, such as physical activity in occupational and and age groups included. When restricted to only large-scale, transportation domains, and of access to specific food types. nationally representative studies, the analysis resulted in For example, people living in rural areas often have higher associations very similar to those presented using the entire levels of physical activity, reflecting their agricultural dataset; heterogeneity decreased after excluding all countries occupations and the need to walk longer distances for day- that did not fulfill these more stringent criteria. to-day activities [15]. Similarly, rural and urban populations may have differential access to various food types, possibly Statistical Analysis with seasonal variations. Urban populations are also likely to A local regression model was used to estimate the income– have higher access to screening and treatment for risks such risk relationships [20]. A local regression model estimates the as high blood pressure and cholesterol. association across income levels without assuming a para- metric model. Rather, the data determine the fitted curve. Data Sources This is a preferable approach when there is no theoretical Data sources for risk factors and economic indicators are model for the shape of the association. provided in Table 1 (see Table S1 for list of countries). Systolic blood pressure (SBP), cholesterol, and BMI were age- Results standardized so that results could be compared across populations with different age structures. We used the World When considered in relation to national income, mean Health Organization (WHO) standard population [16] be- population BMI and cholesterol increased, then flattened, cause it is a better representation of current population age and eventually declined (Figures 2 and 3). Mean BMI structures than older standard populations (e.g., SEGI). The increased most rapidly until a national income of about data used in the analysis were collected by researchers using I$5,000 (see Table S1 for national incomes of individual different instruments. Strict criteria for study selection were countries) and peaked at about I$12,500 for females and applied to ensure methodological rigor and representative- I$17,000 for males. Cholesterol’s point of inflection and peak

Figure 2. Pair-Wise Relationships of Mean Population BMI, SBP, and Total Cholesterol with National Income, Share of Household Expenditure Spent on Food, and Proportion of Population in Urban Areas Data for (A) males and (B) females are shown. National income is measured as per-capita gross domestic product (GDP). BHS, Bahamas; CUB, Cuba; EST, Estonia; ETH, Ethiopia; FIN, Finland; GEO, Georigia; GMB, Gambia; IDN, Indonesia; JOR, Jordan; JPN, Japan; KEN, Kenya; KOR, Korea; KWT, Kuwait; MLT, Malta; MWI, Malawi; NGA, Nigeria; NOR, Norway; NPL, Nepal; PNG, Papua New Guinea; POL, Poland; RUS, Russian Federation; SAU, Saudi Arabia; SLB, Solomon Islands; THA, Thailand; TJK, Tajikistan; TZA, Tanzania; USA, United States; VNM, Viet Nam; WSM, Samoa; ZWE, Zimbabwe. DOI: 10.1371/journal.pmed.0020133.g002

PLoS Medicine | www.plosmedicine.org0406 May 2005 | Volume 2 | Issue 5 | e133 Global Patterns of Nutritional Risks

PLoS Medicine | www.plosmedicine.org0407 May 2005 | Volume 2 | Issue 5 | e133 Global Patterns of Nutritional Risks

(about I$8,000 and I$18,000, respectively) were at higher 2 and 3). Mean population SBP was either not correlated or income levels than those for BMI. The BMI decline relative to only very weakly correlated with cholesterol, BMI, and the the peak at high incomes was larger for females than for economic variables. High SBP levels were observed in some males. The lower mean female BMI at high-income levels developing countries (e.g., Gambia, Ghana, and Nigeria) that (except in the United States where the female BMI was 28.7 had low mean cholesterol or BMI. Similarly, mean population kg/m2) is consistent with the evidence on declining female SBP varied by 20 mm Hg in the countries of Western Europe, BMI in some high-income countries (e.g., Japan [21]) over which have relatively similar income. time, and with the inverse relationship between social class Individual-level epidemiological studies have established a (as measured by education) and female obesity within upper- relationship between BMI and blood pressure [24,25]. Blood middle-income countries [11]. pressure is, however, affected by other aspects of diet, Figures 2 and 3 show that in countries where food is a including salt and fruits and vegetables [26–30] that may vary relatively small proportion of total household expenditure across populations independently of BMI, based on cultural (less than 30%–40%) (i.e., little or no constraints), there was and environmental factors such as food preservation techni- little or no relationship between this factor and mean BMI. ques and subsistence versus commercial access to agricultural Where food is a larger proportion of household expenditure products. Even in within-country studies, the relationships (more than 30%–40%), there was an inverse relationship between blood pressure and or the relative levels of between mean BMI and the proportion of household rural and urban blood pressure have varied, and even expenditure spent on food. Exceptions to this pattern were reversed, in different countries [31–37]. the Russian Federation and Tajikistan, with a large food share Multi-variable regression analysis (results not shown) of household expenditure but with relatively high BMI. The confirmed the absence of a relationship between blood departure from the overall pattern by these two countries pressure and the economic variables considered (the coef- may reflect the economic consequences of the collapse of the ficients for SBP were not statistically significant). BMI and former Soviet Union, which have forced households to devote cholesterol both increased significantly with income and as a large share of their expenditure to food without changing the share of expenditure used for food declined, although the their life style determinants of weight (i.e., diet and physical size of the association was attenuated after adjustment for activity). The relationship between the food share of house- other factors. Countries with a larger urban population had hold expenditure and mean blood cholesterol was similar to higher BMI even after adjustment for their income. that for BMI, but, as in the relationship with income, the flattening of the relationship occurred later (i.e., at lower Discussion food shares) for cholesterol than for BMI. The rapid rise in mean population BMI with increasing income indicates Cardiovascular diseases have multiple well-established higher consumption of energy as food purchase constraints behavioral, environmental, and physiological determinants. at very low income levels are removed. That cholesterol Less well-established are the patterns of these risks at the increases more slowly and continues at higher income levels population level in relation to one another and to economic and lower food share of income than BMI may indicate that variables such as income. Knowledge of these patterns is additional income is first used to increase caloric intake, necessary, however, for better design of long-term policies followed by dietary change beyond higher calories (e.g., and interventions that aim to reduce multiple risks that affect switch from unsaturated vegetable fats and oils to animal fat a common set of diseases in different populations, and for [22]) as income rises further. assessing the global health inequality dimensions of cardio- The proportion of the population in urban areas was vascular disease risks. Using data from over 100 countries, we positively correlated with mean BMI and cholesterol (Figures found that BMI increased rapidly, then flattened, and 2 and 3). The relationships of mean BMI and cholesterol with eventually declined with increasing national income. Choles- urbanization also showed some flattening, but this was less terol had a similar, but more delayed, relationship with noticeable than seen in relation to income, and occurred in national income. Mean population blood pressure was not countries with more than 60% of the population in urban correlated or only weakly correlated with the economic areas. Urban living—which alters transportation and occupa- factors considered, or with cholesterol and BMI. The rapid tional patterns as well as access to various foods—may affect rise in BMI with income, and the lack of a relationship for nutrition and activity, and increase population BMI and blood pressure, illustrate that, currently, major transforma- cholesterol, over and above its impacts mediated through tions in patterns of cardiovascular risks occur at much earlier income. Once a country is primarily urbanized, its infra- stages of economic development than implied by the structure development reduces the urban–rural differences ‘‘diseases of affluence’’ paradigm. If current income–risk in access to food and technology [23], leading to the observed relationships observed in Figures 2 and 3 hold as economies flattening of the curve at high urbanization levels. Exceptions grow, rapidly increasing BMI, coupled with the presence of to the relationship between cholesterol and urbanization elevated blood pressure at all income levels, will increasingly were Indonesia and Nigeria (proportion urban 40% or more, concentrate two major cardiovascular risk factors (blood but low cholesterol levels of 3.2–3.6 mmol/l). Both countries pressure and obesity) in populations with currently low have large populations and, despite the formation of urban income levels, and all three risks in currently middle-income centers, still have predominantly agricultural economies and countries. associated lifestyles. The current income–risk relationships shown in Figures 2 At the aggregate level, the relationships of mean BMI and and 3, however, only partially illustrate the potential future cholesterol with each other or with the economic variables magnitude and global distribution of cardiovascular risks. were qualitatively different from those of mean SBP (Figures There are reasons to suspect that the cross-sectional relation-

PLoS Medicine | www.plosmedicine.org0408 May 2005 | Volume 2 | Issue 5 | e133 Global Patterns of Nutritional Risks

Figure 3. Relationship of Mean Population BMI, SBP, and Total Cholesterol with Average National Income, Food Share of Household Expenditure, and Proportion of Population in Urban Areas Relationships were estimated using local regression models applied to the data in Figure 2. Results for (A) males and (B) females are shown. National income was measured as gross domestic product (GDP). The following outlier countries were dropped (see also Results): United States for males and females in the income–BMI relationship, and Russian Federation and Tajikistan for males and females in the food share of household expenditure–BMI relationship. DOI: 10.1371/journal.pmed.0020133.g003 ships shown in Figures 2 and 3 may be changing over time, income societies, the three risk factors will become increas- with important implications for the epidemiological tran- ingly concentrated in low-income and middle-income nations sition. Longitudinal data from the United States and a small relative to high-income countries. number of other countries show an upward shift in the entire In additional to nutritional risks, tobacco smoking is also income–BMI relationship (Figure 4). At the same time, a an important risk factor for cardiovascular diseases. Cur- downward shift in the relationship between blood pressure rently, an estimated 930 million of the world’s 1.1 billion and cholesterol with income occurred during the 1980s, smokers live in the developing world [42]. Tobacco smoking followed by stabilization or a slight increase in the 1990s for increased among men, followed by women, in industrialized blood pressure, and continued decline for cholesterol (Figure nations in the last century, and has subsequently declined in 4) [38,39]. Definitive explanations for these trends are not some nations (e.g., Canada, the United States, and the United available. Possible reasons for BMI increase include system- Kingdom) [43]. Descriptive models based on historical atic changes in diet and physical activity due to increased patterns in the industrialized world [44] would predict a access to private transportation, television, and manufac- decline in male smoking and an increase in female smoking in tured/packaged foods as a result of technological change, the developing world over the coming decades. However, urbanization, and organization of work [40,41]. For blood there have been major recent transformations in global pressure and cholesterol, possible contributors to the initial tobacco trade, marketing, and regulatory control. As a result, decline include changes in diet (increased access to fresh tobacco consumption among men and women in most fruits and vegetables, and lower salt intake) and use of nations is primarily determined by opposing industry efforts pharmacological interventions (anti-hypertensives and sta- and tobacco control measures [45], and by the socio-cultural tins); increasing obesity may have been the obstacle to further context, rather than national income. Sex-specific data on blood pressure decline. If a similar upward shift in the tobacco consumption are very rare; the available data on income–BMI relationship occurs globally, overweight and smoking prevalence [42,46] indicate little relationship with obesity will play an even larger role in disease burden in national income. developing countries, because these countries will be on an The observed patterns of multiple nutritional risk factors even higher income trajectory of BMI than shown in Figure 3. also have implications for risk factor measurement and As interventions for blood pressure and cholesterol are surveillance. Ideally, analysis of cardiovascular disease risks adopted, and exposure to these risks thus is lowered, in high- should use longitudinal data to assess the effects of socio-

PLoS Medicine | www.plosmedicine.org0409 May 2005 | Volume 2 | Issue 5 | e133 Global Patterns of Nutritional Risks

Figure 4. Shifting Relationships of BMI, SBP, and Total Cholesterol with Income in the United States, Estimated Using Local Regression Data are from the National Health and Examination Survey, 1976–1980, 1988–1992, and 1999–2000. DOI: 10.1371/journal.pmed.0020133.g004 economic factors on nutrition and nutritional risk factors are also important because they cause a number of non- both within and across populations. Detailed longitudinal cardiovascular outcomes (cancers, diabetes, and osteoarthri- data on nutrition (including size and composition of diet and tis) which cannot be addressed by reducing risk factors such frequency of eating) and physical activity [47,48], or on as blood pressure and cholesterol. Current intervention associated indicator risk factors like blood pressure and options for obesity in principle include those that reduce cholesterol, are very scarce, especially in developing coun- caloric intake (e.g., agriculture and food policy and pricing) tries. Partly because of data limitations, the emphasis of and those that increase energy expenditure (e.g., urban global nutritional surveillance has been on dietary composi- design and transportation) [14,40]. There is currently limited tion using aggregate date sources (e.g., the United Nations evidence on the community effectiveness of these interven- Food and Agriculture Organization’s FAOSTAT; http://apps.- tions [52,53]. This limitation highlights the need for research fao.org/). Analyses of these aggregate data sources indicate on design and evaluation of interventions for obesity—which that a shift towards ‘‘Western diets’’ high in saturated fat and in turn requires better data on the relative contributions of sugar and low in fiber is occurring [2,41]. Food composition nutritional and physical activity to the current trends in and total caloric intake, although interrelated [49,50], would weight gain—and on the obesity implications of policies and have distinct influences on intermediate risks like BMI and programs in sectors like transportation and agriculture. cholesterol. The observed rapid BMI rise with national Multi-risk assessment also demonstrates that personal and income indicates that dietary composition, which dominates population-level interventions for blood pressure and cho- global nutritional surveillance, should also be explicitly lesterol (e.g., salt awareness and regulation, and pharmaco- considered in relation to other determinants of weight such logical interventions) should be pursued together with as frequency of eating and physical activity in leisure, attempts to curb or reduce obesity, because these interven- occupation, and transportation domains [51]. Better data tions are effective in reducing the cardiovascular consequen- on food consumption and physical activity in turn require ces of nutritional risk factors [54,55]. developing low-cost and valid instruments for large-scale The division between the and affluence population health surveys. has provided a convenient tool for targeting policies towards Multi-risk assessment also demonstrates intervention needs risks such as undernutrition that affect the poor [56]. and options in societies at different income levels. The Demographic and technological change, however, are in- observed rapid BMI increase with national income indicates creasingly modifying the income patterns of cardiovascular that preventing obesity, which may be more effective than risk factors and shifting their burden to the developing world. reacting after it has occurred [52], should be a priority during As a result, low-income and middle-income countries economic growth and urbanization. Overweight and obesity increasingly face the double burden of infectious disease

PLoS Medicine | www.plosmedicine.org0410 May 2005 | Volume 2 | Issue 5 | e133 Global Patterns of Nutritional Risks and cardiovascular risk factors. Unless the research and quantification of health risks: Global and regional burden of disease attributable to selected major risk factors. Geneva: World Health intervention needs described earlier are pursued, this will Organization. pp. 391–496. create a world in which all major diseases are the diseases of 19. James WTP, Leach R, Mhurchu CN, Kalamara E, Shayeghi M, et al. (2004) the poor. Overweight and obesity (high body mass index). In: Ezzati M, Lopez AD, Rodgers A, Murray CJL, editors. Comparative quantification of health risks: Global and regional burden of disease attributable to selected major risk factors. Geneva: World Health Organization. pp. 497–596. Supporting Information 20. Cleveland WS, Devlin SJ (1988) Locally-weighted regression: An approach to regression analysis by local fitting. J Am Stat Assoc 83: 596–610. 21. Yoshiike N, Seino F, Tajima S, Arai Y, Kawano M, et al. (2002) Twenty-year Table S1. Countries Used in the Analysis and Data Availability changes in the prevalence of overweight in Japanese adults: The National Found at DOI: 10.1371/journal.pmed.0020133.st001 (38 KB PDF). Nutrition Survey 1976–1995. Obes Rev 3: 183–190. 22. Popkin BM (2002) The shift in stages of the nutrition transition in the developing world differs from past experiences. Public Health Nutr 5: 205– 214. Acknowledgments 23. Popkin BM (1999) Urbanization, lifestyle changes and the nutrition This work was sponsored by the National Institute on Aging (grants transition. World Dev 27: 1905–1916. 24. Mertens IL, Van Gaal LF (2000) Overweight, obesity, and blood pressure: PO1-AG17625 and 1-P30-AG024409). The sponsor had no role in The effects of modest weight reduction. Obes Res 8: 270–278. study design, data collection and analysis, decision to publish, or 25. Whelton PK, Appel LJ, Espeland MA, Applegate WB, Ettinger WH Jr, et al. preparation of the manuscript. We thank Xu Ke for data to estimate (1998) Sodium reduction and weight loss in the treatment of the food share of household expenditure, Barry Popkin for in older persons: A randomized controlled trial of nonpharmacologic information about BMI data sources, and Walter Willet for comments interventions in the elderly (TONE). TONE Collaborative Research Group. on an earlier version of this manuscript. & JAMA 279: 839–846. 26. Law MR, Frost CD, Wald NJ (1991) By how much does dietary salt reduction References lower blood pressure. BMJ 302: 811–824. 1. Murray CJL, Lopez AD, editors (1996) The global burden of disease: A 27. Sacks FM, Svetkey LP, Vollmer WM, Appel LJ, Bray GA, et al. (2001) Effects comprehensive assessment of mortality and disability from diseases, on blood pressure of reduced dietary sodium and the Dietary Approaches injuries, and risk factors in 1990 and projected to 2020. Cambridge to Stop Hypertension (DASH) diet. DASH-Sodium Collaborative Research (Massachusetts): Harvard School of Public Health. 990 p. Group. N Engl J Med 344: 3–10. 2. Reddy KS, Yusuf S (1998) Emerging epidemic of cardiovascular disease in 28. Elliott P, Stamler J, Nichols R, Dyer AR, Stamler R, et al. (1996) Intersalt developing countries. Circulation 97: 596–601. revisited: Further analyses of 24 hour sodium excretion and blood pressure 3. Yusuf S, Reddy S, Ounpuu S, Anand S (2001) Global burden of within and across populations. Intersalt Cooperative Research Group. BMJ cardiovascular diseases: Part I: General considerations, the epidemiological 312: 1249–1253. transition, risk factors, and impact of urbanization. Circulation 104: 2746– 29. Lampe J (1999) Health effects of vegetables and fruits: Assessing 2753. mechanisms of action in human experimental studies. Am J Clin Nutr 70: 4. Bulatao RA (1993) Mortality by cause, 1970 to 2015. In: Gribble JN, Preston 475S-490S. SH, editors. The epidemiological transition: Policy and planning implica- 30. Appel LJ, Moore TJ, Obarzanek E, Vollmer WM, Svetkey LP, et al. (1997) A tions for developing countries. Washington (D.C.): National Academy Press. clinical trial of the effects of dietary patterns on blood pressure. DASH pp. 42–68. Collaborative Research Group. N Engl J Med 336: 1117–1124. 5. McKeown T (1988) The origins of human disease. Oxford: Blackwell. 233 p. 31. Edwards R, Unwin N, Mugusi F, Whiting D, Rashid S, et al. (2000) 6. Trowell HC, Burkitt DP, editors (1981) Western diseases, their emergence Hypertension prevalence and care in an urban and rural area of Tanzania. J and prevention. Cambridge (Massachusetts): Harvard University Press. 456 Hypertens 18: 145–152. p. 32. Giles WH, Pacque M, Greene BM, Taylor HR, Munoz B, et al. (1994) 7. Olshansky SJ, Ault AB (1986) The fourth stage of the epidemiological Prevalence of hypertension in rural west Africa. Am J Med Sci 308: 271– transition: The age of delayed degenerative diseases. Milbank Mem Fund Q 275. 64: 355–391. 33. Hunter JM, Sparks BT, Mufunda J, Musabayane CT, Sparks HV, et al. (2000) 8. Marmot MG, Shipley MJ, Rose G (1984) Inequalities in death—Specific Economic development and women’s blood pressure: Field evidence from explanations of a general pattern? Lancet 1: 1003–1006. rural Mashonaland, Zimbabwe. Soc Sci Med 50: 773–795. 9. Blakely T, Hales S, Kieft, Wilson N, Woodward A (2004) Distribution of risk 34. Lisk DR, Williams DE, Slattery J (1999) Blood pressure and hypertension in factors by poverty. In: Ezzati M, Lopez AD, Rodgers A, Murray CJL, editors. rural and urban Sierra Leoneans. Ethn Dis 9: 254–263. Comparative quantification of health risks: Global and regional burden of 35. M’Buyamba-Kabangu JR, Fagard R, Staessen J, Lijnen P, Amery A (1987) disease attributable to selected major risk factors. Geneva: World Health Correlates of blood pressure in rural and urban Zaire. J Hypertens 5: 371– Organization. pp. 1941–2127. 375. 10. Cooper R, Cutler J, Desvigne-Nickens P, Fortmann S, Friedman L, et al. 36. Swai AB, McLarty DG, Kitange HM, Kilima PM, Tatalla S, et al. (1993) Low (2000) Trends and disparities in coronary heart disease, stroke, and other prevalence of risk factors for coronary heart disease in rural Tanzania. Int J cardiovascular diseases in the United States: Findings of the national Epidemiol 22: 651–659. conference on cardiovascular disease prevention. Circulation 102: 3137– 37. van der Sande MA, Bailey R, Faal H, Banya WA, Dolin P, et al. (1997) 3147. Nationwide prevalence study of hypertension and related non-communi- 11. Monteiro CA, Moura EC, Conde WL, Popkin BM (2004) Socioeconomic cable diseases in The Gambia. Trop Med Int Health 2: 1039–1048. status and obesity in adult populations of developing countries: A review. 38. Hajjar I, Kotchen TA (2003) Trends in prevalence, awareness, treatment, Bull World Health Organ 82: 940–946. and control of hypertension in the United States, 1988–2000. JAMA 290: 12. Beaglehole R, Yach D (2003) Globalisation and the prevention and control 199–206. of non-communicable disease: The neglected chronic diseases of adults. 39. Ford ES, Mokdad AH, Giles WH, Mensah GA (2003) Serum total cholesterol Lancet 362: 903–908. concentrations and awareness, treatment, and control of hypercholester- 13. Yach D, Hawkes C, Gould CL, Hofman KJ (2004) The global burden of olemia among US adults: Findings from the National Health and Nutrition chronic diseases: Overcoming impediments to prevention and control. Examination Survey, 1999 to 2000. Circulation 107: 2185–2189. JAMA 291: 2616–2622. 40. Hill JO, Wyatt HR, Reed GW, Peters JC (2003) Obesity and the environ- 14. Drewnowski A, Specter SE (2004) Poverty and obesity: The role of energy ment: Where do we go from here? Science 299: 853–855. density and energy costs. Am J Clin Nutr 79: 6–16. 41. Popkin BM (2002) An overview on the nutrition transition and its health 15. Levine JA, Weisell R, Chevassus S, Martinez CD, Burlingame B, et al. (2001) implications: The Bellagio Meeting. Public Health Nutr 5: 93–103. The work burden of women. Science 294: 812. 42. Jha P, Ranson MK, Nguyen SN, Yach D (2002) Estimates of global and 16. Ahmad OB, Boschi-Pinto C, Lopez AD, Murray CJL, Lozano R, et al. (2000) regional smoking prevalence in 1995, by age and sex. Am J Public Health Age standardization of rates: A new WHO standard. GPE Discussion Paper 92: 1002–1006. No. 31. Geneva: Global Programme on Evidence for Health Policy, World 43. World Health Organization (1997) Tobacco or health: A global status Health Organization. Available:http://w3.whosea.org/healthreport/pdf/pa- report. Geneva: World Health Organization. 504 p. per31.pdf. Accessed 25 March 2005. 44. Lopez AD, Collishaw NE, Piha T (1994) A descriptive model of the cigarette 17. Lawes CMM, Vander Hoorn S, Law MR, Elliott P, MacMahon S, et al. (2004) epidemic in developed countries. Tob Control 3: 242–247. High blood pressure. In: Ezzati M, Lopez AD, Rodgers A, Murray CJL, 45. De Beyer J, Brigden LW, editors (2003) Tobacco control policy: Strategies, editors. Comparative quantification of health risks: Global and regional successes, and setbacks. Washington (D.C.): World Bank. 189 p. burden of disease attributable to selected major risk factors. Geneva: World 46. Shafey O, Dolwick S, Guindon G, editors (2003) Tobacco control country Health Organization. pp. 281–389. profiles 2003. Atlanta (Georgia): American Cancer Society. Available: http:// 18. Lawes CMM, Vander Hoorn S, Law MR, Rodgers A (2004) High cholesterol. www.cancer.org/docroot/PRO/content/PRO_1_1_Tobacco_Control_ In: Ezzati M, Lopez AD, Rodgers A, Murray CJL, editors. Comparative Country_Profiles.asp. Accessed 25 March 2005.

PLoS Medicine | www.plosmedicine.org0411 May 2005 | Volume 2 | Issue 5 | e133 Global Patterns of Nutritional Risks

47. Lock K, Pomerleau J, Causer L, McKee M (2004) Low fruit and vegetable 62. United Nations Population Division (1998) World population prospects: consumption. In: Ezzati M, Lopez AD, Rodgers A, Murray CJL, editors. The 1998 revision. New York: United Nations. 2 v. Comparative quantification of health risks: Global and regional burden of disease attributable to selected major risk factors. Geneva: World Health Organization. pp. 597–728. 48. Bull FC, Armstrong TP, Dixon T, Ham S, Neiman A, et al. (2004) Physical Patient Summary inactivity. In: Ezzati M, Lopez AD, Rodgers A, Murray CJL, editors. Comparative quantification of health risks: Global and regional burden of Background The pattern of cardiovascular diseases is influenced by disease attributable to selected major risk factors. Geneva: World Health many lifestyle factors such as diet, physical activity, work and leisure, and Organization. pp. 729–881. smoking. The effects of these factors are partly mediated through 49. Prentice AM, Jebb SA (2003) Fast foods, energy density and obesity: A intermediate risk factors like overweight and obesity, blood pressure, possible mechanistic link. Obes Rev 4: 187–194. and cholesterol. As societies grow richer, the patterns of risk-factor 50. Blundell JE, Macdiarmid JI (1997) Passive overconsumption. Fat intake and exposure change. Understanding these changes is important for health short-term energy balance. Ann N Y Acad Sci 827: 392–407. policies and interventions. 51. Prentice A, Jebb S (2004) Energy intake/physical activity interactions in the homeostasis of body weight regulation. Nutr Rev 62: S98–S104. What Did the Researchers Do? They looked at cardiovascular disease 52. Friedman JM (2004) Modern science versus the stigma of obesity. Nat Med risks such as being overweight or obese, systolic blood pressure, and 10: 563–569. total cholesterol, and related them to national income, food purchase 53. Kopelman PG (2000) Obesity as a medical problem. Nature 404: 635–643. constraints, and urbanization. Body mass index (BMI) and cholesterol 54. Murray CJL, Lauer JA, Hutubessy RC, Niessen L, Tomijima N, et al. (2003) Effectiveness and costs of interventions to lower systolic blood pressure increased as national income increased, then flattened, and eventually and cholesterol: A global and regional analysis on reduction of cardiovas- declined. BMI increased until an income of about I$12,500 (international cular-disease risk. Lancet 361: 717–725. dollars—a currency adjusted for differences in prices in different 55. Yusuf S, Reddy S, Ounpuu S, Anand S (2001) Global burden of countries) for females and I$17,000 for males, then flattened, and cardiovascular diseases: Part II: Variations in cardiovascular disease by eventually declined. Cholesterol showed the same pattern, but with specific ethnic groups and geographic regions and prevention strategies. some delay. As the proportion of household spending devoted to food Circulation 104: 2855–2864. decreased (i.e., due to changes in income or the price of food), BMI and 56. Gwatkin DR, Guillot M, Heuveline P (1999) The burden of disease among cholesterol levels increased. Also, as more people lived in cities, the the global poor. Lancet 354: 586–589. population’s BMI went up. 57. Ezzati M, Lopez AD, Rodgers A, Vander Hoorn S, Murray CJL, et al. (2002) Selected major risk factors and global and regional burden of disease. What Do These Findings Mean? These findings are from a comparison Lancet 360: 1347–1360. 58. Ezzati M, Vander Hoorn S, Rodgers A, Lopez AD, Mathers CD, et al. (2003) of countries at one point in time, so they may only partly predict the Estimates of global and regional potential health gains from reducing path of individual countries over time. What they do suggest is that multiple major risk factors. Lancet 362: 271–280. cardiovascular disease risks will increasingly be concentrated in low- 59. Strong K, Bonita R (2003) The SuRF Report 1. Surveillance of risk factors income and middle-income countries. Therefore, preventing obesity related to noncommunicable diseases: Current status of global data. should now be considered a priority even in these countries, along with Geneva: World Health Organization. Available: http://whqlibdoc.who.int/ measures to control blood pressure, cholesterol, and tobacco use. publications/2003/9241580305.pdf. Accessed 25 March 2005. 60. Heston A, Summers R, Aten B (2002) Penn World Table, Version 6.1 Where Can I Get More Information? The World Health Organization [computer program]. Philadelphia: Center for International Comparisons has a Web site on nutrition: http://www.who.int/nut/ at the University of Pennsylvania. MedlinePlus has a selection of topics on obesity: http://www.nlm.nih. 61. Xu K, Evans D, Kawabata K, Zeramdini R, Klavus J, et al. (2003) Household gov/medlineplus/obesity.html catastrophic health expenditure: A multicountry analysis. Lancet 362: 111– 117.

PLoS Medicine | www.plosmedicine.org0412 May 2005 | Volume 2 | Issue 5 | e133